We know it because it is given in the problem description, which you violate if the predictor ‘can “err” as much as it wants’.
We know it because it is given in the problem description, which you violate if the predictor ‘can “err” as much as it wants’.